Convert inductor jobs to Linux Amazon 2023 (pytorch#133352)

A continuation of the migration started in - pytorch#131250 Pull Request resolved: pytorch#133352 Approved by: https://github.com/zxiiro, https://github.com/seemethere
xuanzhang816 · Aug 14, 2024 · 118b2a4 · 118b2a4
1 parent 62cd065
commit 118b2a4
Show file tree

Hide file tree

Showing 2 changed files with 80 additions and 60 deletions.
diff --git a/.github/workflows/_linux-build.yml b/.github/workflows/_linux-build.yml
@@ -34,6 +34,11 @@ on:
         default: "5.2"
         description: |
           List of CUDA architectures CI build should target.
+      runner_prefix:
+        required: false
+        default: ""
+        type: string
+        description: Prefix for runner label
       runner:
         required: false
         type: string
@@ -91,7 +96,7 @@ jobs:
   build:
     # Don't run on forked repos
     if: github.repository_owner == 'pytorch'
-    runs-on: ${{ inputs.runner }}
+    runs-on: ${{ inputs.runner_prefix}}${{ inputs.runner }}
     timeout-minutes: 240
     outputs:
       docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}

diff --git a/.github/workflows/inductor.yml b/.github/workflows/inductor.yml
@@ -18,34 +18,44 @@ concurrency:
 permissions: read-all
 
 jobs:
+  get-label-type:
+    name: get-label-type
+    uses: ./.github/workflows/_runner-determinator.yml
+    with:
+      triggering_actor: ${{ github.triggering_actor }}
+      issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
+      curr_branch: ${{ github.head_ref || github.ref_name }}
+
   linux-focal-cuda12_1-py3_10-gcc9-inductor-build:
     name: cuda12.1-py3.10-gcc9-sm86
     uses: ./.github/workflows/_linux-build.yml
+    needs: get-label-type
     with:
       build-environment: linux-focal-cuda12.1-py3.10-gcc9-sm86
       docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9-inductor-benchmarks
       cuda-arch-list: '8.6'
+      runner_prefix: "amz2023."
       test-matrix: |
         { include: [
-          { config: "inductor", shard: 1, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "inductor", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "inductor_distributed", shard: 1, num_shards: 1, runner: "linux.g5.12xlarge.nvidia.gpu" },
-          { config: "inductor_huggingface", shard: 1, num_shards: 1, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "inductor_timm", shard: 1, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "inductor_timm", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "inductor_torchbench", shard: 1, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "inductor_torchbench", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "dynamic_inductor_huggingface", shard: 1, num_shards: 1, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "dynamic_inductor_timm", shard: 1, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "dynamic_inductor_timm", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "dynamic_inductor_torchbench", shard: 1, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "dynamic_inductor_torchbench", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "aot_inductor_huggingface", shard: 1, num_shards: 1, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "aot_inductor_timm", shard: 1, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "aot_inductor_timm", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "aot_inductor_torchbench", shard: 1, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "aot_inductor_torchbench", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "inductor_cpp_wrapper_abi_compatible", shard: 1, num_shards: 1, runner: "linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor", shard: 1, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor", shard: 2, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor_distributed", shard: 1, num_shards: 1, runner: "amz2023.linux.g5.12xlarge.nvidia.gpu" },
+          { config: "inductor_huggingface", shard: 1, num_shards: 1, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor_timm", shard: 1, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor_timm", shard: 2, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor_torchbench", shard: 1, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor_torchbench", shard: 2, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "dynamic_inductor_huggingface", shard: 1, num_shards: 1, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "dynamic_inductor_timm", shard: 1, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "dynamic_inductor_timm", shard: 2, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "dynamic_inductor_torchbench", shard: 1, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "dynamic_inductor_torchbench", shard: 2, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "aot_inductor_huggingface", shard: 1, num_shards: 1, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "aot_inductor_timm", shard: 1, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "aot_inductor_timm", shard: 2, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "aot_inductor_torchbench", shard: 1, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "aot_inductor_torchbench", shard: 2, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor_cpp_wrapper_abi_compatible", shard: 1, num_shards: 1, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
         ]}
     secrets:
       HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
@@ -64,14 +74,16 @@ jobs:
   linux-focal-cuda12_1-py3_12-gcc9-inductor-build:
     name: cuda12.1-py3.12-gcc9-sm86
     uses: ./.github/workflows/_linux-build.yml
+    needs: get-label-type
     with:
       build-environment: linux-focal-cuda12.1-py3.12-gcc9-sm86
       docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3.12-gcc9-inductor-benchmarks
       cuda-arch-list: '8.6'
+      runner_prefix: "amz2023."
       test-matrix: |
         { include: [
-          { config: "inductor", shard: 1, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "inductor", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor", shard: 1, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor", shard: 2, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
         ]}
 
   linux-focal-cuda12_1-py3_12-gcc9-inductor-test:
@@ -89,9 +101,10 @@ jobs:
     with:
       build-environment: linux-jammy-py3.12-gcc11
       docker-image-name: pytorch-linux-jammy-py3.12-halide
+      runner_prefix: "amz2023."
       test-matrix: |
         { include: [
-          { config: "inductor-halide", shard: 1, num_shards: 1, runner: "linux.12xlarge" },
+          { config: "inductor-halide", shard: 1, num_shards: 1, runner: "amz2023.linux.12xlarge" },
         ]}
 
   linux-jammy-cpu-py3_12-inductor-halide-test:
@@ -112,10 +125,11 @@ jobs:
       build-environment: linux-focal-cuda12.4-py3.10-gcc9-sm86
       docker-image-name: pytorch-linux-focal-cuda12.4-cudnn9-py3-gcc9-inductor-benchmarks
       cuda-arch-list: '8.6'
+      runner_prefix: "amz2023."
       test-matrix: |
         { include: [
-          { config: "inductor_timm", shard: 1, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "inductor_timm", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor_timm", shard: 1, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor_timm", shard: 2, num_shards: 2, runner: "amz2023.linux.g5.4xlarge.nvidia.gpu" },
         ]}
     secrets:
       HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
@@ -138,43 +152,44 @@ jobs:
     with:
       build-environment: linux-jammy-py3.8-gcc11-build
       docker-image-name: pytorch-linux-jammy-py3.8-gcc11-inductor-benchmarks
+      runner_prefix: "amz2023."
       test-matrix: |
         { include: [
-          { config: "inductor_avx512", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "inductor_avx512", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_huggingface", shard: 1, num_shards: 1, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_timm", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_timm", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_torchbench", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_torchbench", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_huggingface_freezing", shard: 1, num_shards: 1, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_timm_freezing", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_timm_freezing", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_torchbench_freezing", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_torchbench_freezing", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_huggingface_amp_freezing", shard: 1, num_shards: 1, runner: "linux.16xlarge.spr" },
-          { config: "cpu_inductor_timm_amp_freezing", shard: 1, num_shards: 2, runner: "linux.16xlarge.spr" },
-          { config: "cpu_inductor_timm_amp_freezing", shard: 2, num_shards: 2, runner: "linux.16xlarge.spr" },
-          { config: "cpu_inductor_torchbench_amp_freezing", shard: 1, num_shards: 2, runner: "linux.16xlarge.spr" },
-          { config: "cpu_inductor_torchbench_amp_freezing", shard: 2, num_shards: 2, runner: "linux.16xlarge.spr" },
-          { config: "dynamic_cpu_inductor_huggingface", shard: 1, num_shards: 1, runner: "linux.12xlarge" },
-          { config: "dynamic_cpu_inductor_timm", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "dynamic_cpu_inductor_timm", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "dynamic_cpu_inductor_torchbench", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "dynamic_cpu_inductor_torchbench", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_aot_inductor_huggingface_freezing", shard: 1, num_shards: 1, runner: "linux.12xlarge" },
-          { config: "cpu_aot_inductor_timm_freezing", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_aot_inductor_timm_freezing", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_aot_inductor_torchbench_freezing", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_aot_inductor_torchbench_freezing", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "inductor_torchbench_cpu_smoketest_perf", shard: 1, num_shards: 1, runner: "linux.24xl.spr-metal" },
-          { config: "inductor_avx2", shard: 1, num_shards: 2, runner: "linux.10xlarge.avx2" },
-          { config: "inductor_avx2", shard: 2, num_shards: 2, runner: "linux.10xlarge.avx2" },
-          { config: "cpu_inductor_huggingface_freezing_avx2", shard: 1, num_shards: 1, runner: "linux.10xlarge.avx2" },
-          { config: "cpu_inductor_torchbench_freezing_avx2", shard: 1, num_shards: 2, runner: "linux.10xlarge.avx2" },
-          { config: "cpu_inductor_torchbench_freezing_avx2", shard: 2, num_shards: 2, runner: "linux.10xlarge.avx2" },
-          { config: "cpu_inductor_timm_freezing_avx2", shard: 1, num_shards: 2, runner: "linux.10xlarge.avx2" },
-          { config: "cpu_inductor_timm_freezing_avx2", shard: 2, num_shards: 2, runner: "linux.10xlarge.avx2" },
+          { config: "inductor_avx512", shard: 1, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "inductor_avx512", shard: 2, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_inductor_huggingface", shard: 1, num_shards: 1, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_inductor_timm", shard: 1, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_inductor_timm", shard: 2, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_inductor_torchbench", shard: 1, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_inductor_torchbench", shard: 2, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_inductor_huggingface_freezing", shard: 1, num_shards: 1, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_inductor_timm_freezing", shard: 1, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_inductor_timm_freezing", shard: 2, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_inductor_torchbench_freezing", shard: 1, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_inductor_torchbench_freezing", shard: 2, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_inductor_huggingface_amp_freezing", shard: 1, num_shards: 1, runner: "amz2023.linux.16xlarge.spr" },
+          { config: "cpu_inductor_timm_amp_freezing", shard: 1, num_shards: 2, runner: "amz2023.linux.16xlarge.spr" },
+          { config: "cpu_inductor_timm_amp_freezing", shard: 2, num_shards: 2, runner: "amz2023.linux.16xlarge.spr" },
+          { config: "cpu_inductor_torchbench_amp_freezing", shard: 1, num_shards: 2, runner: "amz2023.linux.16xlarge.spr" },
+          { config: "cpu_inductor_torchbench_amp_freezing", shard: 2, num_shards: 2, runner: "amz2023.linux.16xlarge.spr" },
+          { config: "dynamic_cpu_inductor_huggingface", shard: 1, num_shards: 1, runner: "amz2023.linux.12xlarge" },
+          { config: "dynamic_cpu_inductor_timm", shard: 1, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "dynamic_cpu_inductor_timm", shard: 2, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "dynamic_cpu_inductor_torchbench", shard: 1, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "dynamic_cpu_inductor_torchbench", shard: 2, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_aot_inductor_huggingface_freezing", shard: 1, num_shards: 1, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_aot_inductor_timm_freezing", shard: 1, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_aot_inductor_timm_freezing", shard: 2, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_aot_inductor_torchbench_freezing", shard: 1, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "cpu_aot_inductor_torchbench_freezing", shard: 2, num_shards: 2, runner: "amz2023.linux.12xlarge" },
+          { config: "inductor_torchbench_cpu_smoketest_perf", shard: 1, num_shards: 1, runner: "amz2023.linux.24xl.spr-metal" },
+          { config: "inductor_avx2", shard: 1, num_shards: 2, runner: "amz2023.linux.10xlarge.avx2" },
+          { config: "inductor_avx2", shard: 2, num_shards: 2, runner: "amz2023.linux.10xlarge.avx2" },
+          { config: "cpu_inductor_huggingface_freezing_avx2", shard: 1, num_shards: 1, runner: "amz2023.linux.10xlarge.avx2" },
+          { config: "cpu_inductor_torchbench_freezing_avx2", shard: 1, num_shards: 2, runner: "amz2023.linux.10xlarge.avx2" },
+          { config: "cpu_inductor_torchbench_freezing_avx2", shard: 2, num_shards: 2, runner: "amz2023.linux.10xlarge.avx2" },
+          { config: "cpu_inductor_timm_freezing_avx2", shard: 1, num_shards: 2, runner: "amz2023.linux.10xlarge.avx2" },
+          { config: "cpu_inductor_timm_freezing_avx2", shard: 2, num_shards: 2, runner: "amz2023.linux.10xlarge.avx2" },
         ]}
     secrets:
       HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}